The Snow Telemetry (SNOTEL) network has collected data for snow water equivalent (SWE) across the Western United States since the 1970s. Seasonal forecasting of SWE levels in the region is of particular interest due to the importance of snow to water management, ecology, and hydrology. Most existing models for snow storage are concerned only with the spatial or temporal component, require extensive calibration, or cannot be extended to different scales. We propose a Bayesian hierarchical model for analog forecasting of SWE continuously in both time and space. Our model uses stochastic differential equations and potential functions and we utilize a formal network structure to identify analog SWE trajectories. We demonstrate the utility of the model by forecasting SWE based on data from SNOTEL stations in Colorado, USA.